USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE

advertisement
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Michael Kostiuk, MA Geography
Geospatial Analyst, Ottawa, Ontario, Canada
ABSTRACT:
Remote sensing data and Geographic information systems (GIS) are relatively new and potentially valuable tools for coastal zone
management. This paper examines the effectiveness of using remote sensing data to detect sea level change. Since resolution is such
an important and vital element of spatial digital data for use in geographic information systems, it is important to know how to assess
its quality, accuracy and level of precision. Using remote sensing data to detect sea level change also requires accurate historical
baseline spatial data and knowledge of how the coastline is defined and mapped. Map datum refers to the various locations to which
geographic measurements are referenced. This referencing system is an important item on the list of cartographic components that
help to identify and categorize individual maps. For example, many North American maps have been, or will soon be, converted to a
horizontal map datum known as NAD83. Along with horizontal datum, maps are also referenced to vertical datum. The choice of
vertical and horizontal map datum along with other cartographic elements such as map projection, scale and meta data will determine
to what level of precision coastal change can be accurately measured. This paper will explain how to select the most appropriate
baseline spatial data as well as the type of the remote sensing data that will provide the most reliable results for the detection of seal
level change. Cobscook Bay, Maine was used for two case studies to demonstrate some of these coastal mapping parameters.
1.
INTRODUCTION
Recent improvements in the field of remote sensing now
allows for the acquisition of high resolution images for use in a
wide a variety of applications that require the accurate
determination of geographic location. Any application of
Remote Sensing data that ascertains land use and delineates
features will benefit from higher degrees of accuracy. Often the
interface between one type of land use and another type is
“fuzzy” or imprecise and the boundary line is placed in a
somewhat arbitrary location. This often occurs on soil and
vegetation maps where there is usually a gradually transition
from one type to another rather than the blunt transition that is
indicated by the edge of a polygon. In other cases the interface
between one land use type and another type is much more
apparent such as when a land use of urban development
suddenly stops at a set location and a land use of a greenbelt
then starts. Other places where there are even more dramatic
changes in the interface between land use occur along the
coastal zones of the seas and oceans. At these locations the
distinction between the land-side and the water-side of the
coastal zone is very clear. However, the coastal zone is also a
dynamic environment where due to such factors as coastal
erosion and deposition, tides, storms, biological activities
within coral reefs, volcanism, plate tectonics and intervention
by man have created conditions where the location of the landsea interface or coastline is constantly changing. Lillesand and
Kiefer (1994, p28) define two temporal forms of spatial data
acquisition: “time-critical and time-stable”. Time-stable
measurements are made in conditions that do not (except in
exceptional cases such as natural disasters) experience rapid
changes in position such as with fixed geological formations.
Time-critical measurements are made of areas that experience
constant change such as animal migration, automobile usage
on highways or the tides of the world’s oceans. It is this last
element that is of particular concern when attempting to
measure changes in the sea level. A case study approach is
used in this paper to demonstrate some elements of coastal
mapping.
2.
VERTICAL DATUMS
Since the sea level changes from a maximum high tide to a
maximum low tide a variety of vertical datum have been
created for special purposes by many different types of
organizations around the world. According to the USGS and
NOAA (August, 2002) there are 27 different vertical datums in
use. These different datum are either based around an
Orthometric (geodetic) level which is based on Mean Sea
Level (MSL) or they are based on tidal measures of the high
and low water line at the land and sea interface. The advent of
GPS has added a newer form of vertical datum and since it is
based on 3 points it is referred to as a 3-D datum. Some other
types of datum are Mean Higher High water (MHHW), Mean
High Water (MHL) and Mean Lower Low Water (MLLW).
Mean Lower Low Water level is also referred as Chart datum
for hydrographic charts. Mean Seal Level (MSL) is used by the
USGS as the vertical datum for the production of its
topographic maps. These specific MSL orthometric levels are
known as North American Geodetic Datum 1929 (NGVD) or
the newer North American Vertical datum (NAVD 88). The
choice and use of different vertical datum by various mapping
organizations around the world means that there are many
definitions of where the shoreline is located. Since at least one
reliable cartographic baseline is required to measure sea level
change it is vitally important to select the same vertical datum
for both the reference geographic data and the newly acquired
remote sensing data. Otherwise any attempt at change
detection will be adversely affected by the different vertical
datums. Once a vertical datum has been selected the next step
is to acquire reliable and accurate geographic data of the
shoreline to use as a baseline reference.
3.
SELECTING BASELINE DATA
Spatial data that are being used for spatial analysis for coastal
zone management applications such as detecting sea level
change can come from a wide variety of sources and they may
be used for purposes that the spatial data were not designed to
support. The type of spatial data that are selected should be
matched to the specific coastal zone management task that is to
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
be performed (such as seal level change). For example, smallscale spatial data should be used for the creation of maps for an
atlas, medium scale maps should be used for spatial analysis on
an urban or regional scale, and large-scale maps should be used
for various civil operations such as in road building or for the
construction of bridges. The case studies described for this
paper uses analogue and digital spatial data that has been
obtained from various sources and their various features of
scale, precision, accuracy and resolution were explored by
using an GIS (ArcView) to measure the area and length of the
coastal area of Cobscook Bay, Maine. Spatial data can come in
many forms and it can also cover a wide variety of temporal
periods and geographic areas. It should be kept in mind that an
error or inaccuracy in a paper (analogue) map is often
compounded when it is digitized into spatial data.
3.1 Map Accuracy
Analogue maps generally have two forms of accuracy
standards applied to them. The first mapping standard is the
level of precision at which the geographic features were
measured and recorded, and the second mapping standard
refers to the accuracy level regarding the actual drawing of the
geographic features on the map. For example, the United
States National Map Accuracy Standards that were established
in 1941, and later revised in 1947 state the horizontal accuracy
for maps on publication scales larger than 1:20,000 to be “not
more than 10 percent of the points tested shall be in error by
more than 1/30 inch, measured on the publication scale; for
maps on publication scales of 1:20,000 or smaller, 1/50 inch”
(U.S. Bureau of The Budget, 1947). This specification uses the
printed map to measure how close the drawn line or object
should be to the true position of the geographic feature that is
being depicted on the map. The levels of accuracy that are
applied to the measuring and collection of the geographic
features at the source are done according to a different set of
standards. These standards are based on the scale of the map
and the intended use of the map. For example a 1:24,000 scale
map is required to have at least 90 percent of horizontal points
tested to be accurate to within one-fiftieth of an inch on the
map, which at this scale equate to a horizontal accuracy level
of 40 feet on the ground. Other mapping specifications refer to
how accurate the map is to the actual position on the ground.
The Standards and Specifications of the National Topographic
Data Base of Geomatics Canada, version 3.1, refers to
horizontal accuracy of its 1:50,000 scale maps according to
geometric accuracy as the “difference between the position of
the geometric representation associated with an entity and the
real ground position of the corresponding topographic feature,
as measured with respect to the geodetic network” (Geomatics
Canada, 1997, p. 10). In addition to this requirement, the
horizontal accuracy is further divided into three subclasses that
relate to the population density of the map sheet. This produces
a variable accuracy requirement that increases for populated
urban areas, and decreases for rural and isolated areas. The
horizontal accuracy standards for Geomatics Canada’s
1:50,000 maps are such that it aims to meet the following
accuracy requirements: “i) For urban areas, the circular
horizontal accuracy is 10 meters; ii) For rural areas, the circular
horizontal accuracy is 25 meters and; iii) For isolated areas, the
circular horizontal accuracy is 125 meters” (Geomatics Canada
1997, p. 101). These mapping specifications show that
accuracy levels that can vary even when they are at the same
scale and they have been obtained from the same source.
3.2 Determining the Resolution of Spatial Data
Since resolution is such an important and vital element of
spatial digital data for use in geographic information systems,
it is important to know how to assess its quality, accuracy and
level of precision. The resolution of a map or an image is
defined by the smallest individual feature that can be clearly
identified. It is a relatively straightforward process to check the
spatial resolution on paper maps because there is a fairly
obvious ratio between the map scale and its printed resolution.
The spatial resolution on a map is determined by simply
measuring the smallest printed feature on the map, and then
comparing that value to the scale of the map. According to
Tobler (1988) the smallest mark that a cartographer can make
on a map is “approximately one half millimeter in size” (p.
131). From this observation Tobler reported that the resolution
of a map scale can be quickly and easily determined by the use
of a simple formula. To determine the resolution of a map, the
denominator of the map scale is divided by 1000. This will
produce a value for the detectable size of the map in meters,
and the resolution of the map is obtained by further dividing
the value of the detectable size by the value of 2. This formula
will give a value that corresponds to the smallest size of one
half millimeter that a cartographer is able to both represent and
print a geographic feature on a map. The following table
shows the relationship between map scale, detectable size, and
map resolution. Note: These are the same values as the five
scales that were used for the GIS-based analysis of Cobscook
Bay, Maine (Kostiuk, 2001a, 2001b).
Table 1. Adapted from Tobler (1988, p. 32)
Comparison of map scale to map resolution.
Map Scale
Detectable
Size
Map
Resolution
24000
24 meters
12 meters
70000
70 meters
35 meters
100000
100 meters
50 meters
250000
250 meters
125 meters
1000000
1,000 meters
500 meters
The calculations that produced the above table are also based
on the mathematical rule of the sampling interval (Tobler,
1988).
The sampling rule requires that data should be
collected using a measurement system that is at least one half
the unit of the data that are to be recorded. The reason for this
particular rule of statistics is to prevent the loss of data that can
occur if the geographic feature is located between two points
on a grid. To illustrate the sampling theory Tobler (1988) noted
that to detect the movements of a thunderstorm cell for every
one-kilometer of travel a half-kilometer grid is needed. The
reason that the grid is set to one half kilometer is to prevent the
thunderstorm from passing between two stations, and in doing
so, it would not be recorded. The same sampling rule is applied
in the calculations that produced the above table, where the
value of the detection is twice the size of the value of the map
resolution.
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
4.
TESTING THE ACCURACY AND RELIABILITY
OF THE BASELINE DATA.
For the digital spatial analysis of Cobscook Bay (Case Study
A), a high emphasis was placed on the accurate measurement
of the coastline, and therefore it was important to determine a
minimum acceptable scale that will be useful for that purpose.
Wainwright et al. (1991, p.14) noted: “Water bodies must
possess enough resolution to describe reaches. Coastlines must
have enough detail to describe shoreline units. Reaches may be
as small as a 50 m section of a stream. Shoreline units are
normally larger, but may be as small as a 100 m section of
shoreline”. Wainwright reported that a minimum acceptable
scale to portray the coastline should be 1:40,000 and larger,
and that the minimum scale for watershed mapping should be
1:20,000 (Wainwright et al., 1991).
4.1 Scale Requirements
For the GIS analysis of Cobscook Bay (Case study A), this
1:40,000-scale recommendation was used as the desired
smallest acceptable scale (designated as the baseline scale) to
measure the coastline of Cobscook Bay. A search for digital
spatial data was then necessary to locate any sources of data
that met or exceeded this 1:40,000 requirement. Digital spatial
data that represents a range of map scales and sources were
also located to determine how close, or how accurately they are
able to represent the size and dimensions of the bay.
4.2 Horizontal and Vertical Datum Requirements
The coastline is the geographic center of the coastal zone and it
is also the location from where the coastal zone is defined. The
location of the coastline depends on which datum is used for
the sea level, and whether the High tide, Mean Sea Level, or
the Low tide level is to be used as the point of reference. In the
case of Cobscook Bay, like the Bay of Fundy, there is a large
difference between the Low and High tide levels. The mean
tidal range of Cobscook Bay is 5.7 meters or approximately
18.7 feet, the highest tide that can be observed in the United
States of America (Brooks et al., 1999). Consequently, such a
large tidal range as this produces two very different coastlines
at the high and low range of the tide cycle.
5.
SOURCES OF DIGITAL SPATIAL DATA USED TO
MEASURE COBSCOOK BAY
5.2 Maine Office of GIS Spatial Data
The digital spatial data from the Maine Office of GIS were
downloaded as compressed ArcInfo (GIS made by ESRI)
format files at scales of 1:24,000 and 1:100,000. These data
were digitized and referenced to the Mean High Water (MHW)
line as are shown on USGS 1:24,000 scale quadrangle maps.
The accuracy limits for the 1:24,000 scale data is that “ not
more than 10 percent of the points tested shall be in error by
more than 0.02 inch, measured on the publication scale” for
horizontal accuracy (ground scale), and “that not more than 10
percent of the elevations tested shall be in error more than onehalf the contour interval for vertical accuracy” (USGS, Part 1,
1997, p.1. D-2). For a 1:24,000 scale map, the horizontal
accuracy of 0.02 inch on the map equates to 40 feet or 12.19
meters on the ground. The accuracy limits for the 1:100,000
scale data are that at least 90 percent of points tested are within
0.02 inch of the true position (ground scale) for horizontal
accuracy, and that at least 90 percent of well-defined points
tested should be within one-half contour interval of the correct
value for vertical accuracy (USGS, Part 3, 1997, pp. 3-12, 313). For a 1:100,000-scale map, the horizontal accuracy of
0.02 inch on the map represents 166.7 feet or 50.8 meters in
ground terms.
5.3 The United States Geological Service Coastline
Extractor Spatial Data
The data were in the form of NOAA/NOS Medium Resolution
Digital Vector Shoreline at a scale of 1:70,000. These data
were a portion of a larger data set that covers the entire United
States of America. These data were digitized from NOAA
nautical charts. The other set of spatial data set were a portion
of the World Vector Shoreline, which is at the 1:250,000 scale.
These data can be used for worldwide coverage. Both of these
data sets contain only line information, and since no polygon
features are included with these data sets, only the lengths of
features can be easily measured. For the 1:70,00 data, the
horizontal datum is NAD83, and the vertical datum is
NAVD29 which is based on the mean high or mean higher
high shoreline position that is published on nautical charts. The
spatial resolution of the data is set to a minimum adjacent
vertex spacing of five meters ground distance. The source of
the spatial data is from the master copy of the National Ocean
Service’s coast charts, and they are supposed to meet or exceed
National Map Accuracy Standards (Rohmann, 2000).
5.4 The Results of the Spatial Analysis of Cobscook Bay:
Case Study A
5.1 Digital Spatial Data
To obtain digital spatial data of the study area in a vector
format that meets the requirements that were specified, a
search of the Internet produced the following freely available
sources of data:
1)
Maine
Office
of
GIS
Internet
site
at
http://apollo.ogis.state.me.us
2)
The United States Geological Service Coastline
Extractor at http://crusty.er.usgs.gov/coast/getcoast.html
3)
Digital Chart of the World (DCW) data from the
Pennsylvania State University's Maps Library site at
http://ortelius.maproom.psu.edu/dcw/.
A more detailed description of the larger sets of digital spatial
data is as follows:
The spatial analysis of Cobscook Bay (Table 2) that was based
on the five scale-based sets of digital spatial data clearly show
a wide range of results that can be directly attributed to map
generalization, spatial resolution and map scale. There are also
at least three different opinions on the extent of the bay and
therefore the spatial analysis measured Cobscook Bay three
times for each set of spatial data. Not unexpectedly, the length
of the shoreline of both the mainland and the islands of
Cobscook Bay show decreasing lengths as the scale of the
spatial data becomes smaller. Using the Estes Head to Lubec
town pier limit of Cobscook Bay as an example, the length of
the mainland varies from a value of 380,900 meters using the
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
TABLE 2. DIMENSIONS OF COBSCOOK BAY, MAINE, USA (Kostiuk, 2001a, 2001b).
Scale
24,000
Baseline
Scale
70,000
100,000
250,000
1,000,000
357,856
363,499
380,900
315,816
321,249
335,820
299,445
304,562
317,125
192,493
198,361
220,612
168,850
171,969
180,019
57,482
57,593
60,860
51,674
51,674
55,709
31,786
31,786
35,737
20,884
20,884
25,917
0
0
1,223
169
171
186
87
87
95
29
29
33
10
10
13
0
0
1
89,946,703
92,070,024
100,536,850
N/A
N/A
N/A
90,197,323
92,297,120
100,534,727
N/A
N/A
N/A
94,019,162
95,991,320
105,004,516
2,404,855
2,405,224
2,639,025
N/A
N/A
N/A
1,962,162
1,962,162
2,181,156
N/A
N/A
N/A
0
0
201,450
Length of Mainland shoreline (meters)
Comstock Pt.-Shackford
Comstock Pt.-Estes Hd.
Estes Hd.-Lubec Pier
Length of Island shoreline (meters)
Comstock Pt.-Shackford
Comstock Pt.-Estes Head.
Estes Head.-Lubec Pier
Number of Islands
Comstock Pt.-Shackford
Comstock Pt.-Estes Head.
Estes Head.-Lubec Pier
Area of Cobscook Bay (meters 2)
Comstock Pt.-Shackford
Comstock Pt.-Estes Head.
Estes Head.-Lubec Pier
Area of Islands (meters 2)
Comstock Pt.-Shackford
Comstock Pt.-Estes Head.
Estes Head.-Lubec Pier
1:24,000 scale digital spatial data, to a length of 180,019
meters that is based on the 1:1,000,000 scale digital spatial
data. This is a difference of 200,881 meters or 2.1 times the
length of the shoreline for the 1:24,000 digital spatial data over
the use of the 1:1,000,000 digital spatial data. Based on the
results of table 2 the accurate depiction of the geographic
reality of Cobscook Bay is directly dependent upon the scale of
a particular set of digital spatial data.
6.
USING REMOTE SENSING DATA (CASE STUDY
B): COBSCOOK BAY, MAINE
The purpose of Case Study B was twofold. First it explored the
suitability of using digital remote sensing data (such as
LANDSAT 5 TM satellite data) within a GIS for coastal zone
management applications (as in detecting sea level change).
Digital remote sensing data may be an alternative source for
coastal zone management applications if the needed digital
spatial data are unavailable. And second, it demonstrates the
transition of spatial data into geographic information through
the use of the LANDSAT 5 TM data within a GIS. The data
that are provided from LANDSAT 5 are in a raw format and it
needs to be processed with a GIS so that ground features can
be properly identified. Once the LANDSAT 5-based ground
features are identified and georeferenced they become
geographic information.
7.
USING LANDSAT 5 TM DATA FOR COASTAL
CHANGE DETECTION. CASE STUDY B.
7.1 Establishing the Vertical Datum
A first step to consider when using aerial images and remote
sensing data such as LANDSAT 5 TM data to measure a
coastal zone area is to decide on which vertical datum to use to
define the shoreline. A statement on the vertical datum is not
normally supplied with the LANDSAT 5 data, and some
method must be found to establish the preferred vertical datum
for the coastline. One method of determining the preferred
location of the coastline is to acquire the images when the tide
is at its maximum high tide cycle. This can be done by
checking the tide tables to see when the high tide period is
going to occur. Acquiring aerial images during the high tide
cycle is a method used by mapping organizations to identify
the location of the coastline. The Geomatics division of Natural
Resources Canada’s National Topographic System 1:50,000
Standards and Specifications manual defines the high water
mark (MHW) according to various conditions such as:
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
Tidal waters:
i) The water’s edge at the time of photography, if the tide is
high according to the tide tables of the Canadian
Hydrographic Service.
ii) The demarcation line caused by the change in vegetation
or the deposit of sea debris if the tide is not high according
to the tide tables of the Canadian Hydrographic Service.
iii) The zero contour if the tide is not high and the
demarcation line cannot be seen. (Geomatics Canada,
1997, p. 25/34).
The MHW is used to define the coastline on other topographic
maps such as the USGS 1:24,000 scale-based quadrangle maps
that cover the coastline of the United States of America.
Another method to locate the high tide line on aerial images is
to use historical tide records of the area, and then to locate an
archived aerial image or remote sensing image that was taken
during a peak high tide cycle. It is also much easier and more
practical to compare the LANDSAT digital spatial data to
other forms of digital spatial data if they contain the same
parameters such as map projections, map units and horizontal
and vertical datums.
Using the same cartographic
specifications in both case studies also ensures the spatial
analysis is based on a consistent methodology.
7.2 Locating, Selecting and Obtaining LANDSAT 5 Data
The data were obtained from the EROS Internet site of the
United States Geological Service through a search of the
historical archive of LANDSAT 5 data. Cross referencing the
LANDSAT 5 data to the historical tide data indicated that the
image taken at 14:41:12.1375 GMT on July 21, 1989 was two
hours after the high tide. The metadata for the LANDSAT 5
TM data contained the following information:
•
Sensor: TM (This is a short form for Thematic Mapper
which is one of the internal sensors of LANDSAT 5).
Processing Level: Systematic Geocorrection (The type of
correction performed on the LANDSAT 5 data).
Map Projection: UTM (Universal Transverse Mercator
projection).
UTM Zone: 19 (The number of the UTM zone).
Earth Ellipsoid: WGS84 (A horizontal datum which is
equivalent to NAD 83).
Image Lines: 6299 (The number of lines in the
LANDSAT 5 data file).
Image Pixels: 6864 (The number of pixels in the
LANDSAT 5 data file).
Interleaving: BSQ (This stands for Band Sequential).
•
•
•
•
•
•
•
Since the LANDSAT 5 TM image was not taken when the
water level in Cobscook Bay was at the high tide level, a
method was developed to map the coastline at the MHW level.
The development and application of this method forms the
basis for the spatial analysis of Cobscook Bay in case study B.
8.
METHODOLOGY OF THE SPATIAL ANALYSIS
OF THE CASE STUDY OF COBSCOOK BAY,
PART B.
8.1 Step One: Selecting a System That Can Interpret and
Process the LANDSAT 5 Data
There are various computer programs that allow the viewing
and processing data from remote sensing systems such as PCI
Geomatic’s Image Handler. The PCI program is able to read
the LANDSAT 5 data, and it then converts these data to its
own format for processing and manipulation. The suites of PCI
programs are also used by many governmental and private
organizations to process digital spatial data.
8.2 Step Two: Creating a Subset
A subset covering the study area was created from the large
image and the most suitable spectral bands for delineating the
coastline at the MHW (Mean High Water) level were selected.
An explanation of the properties of the preferred spectral bands
are shown in step three.
8.3 Step Three: Identifying Separate Areas of the Coastal
Zone.
PCI’s Image Handler can display LANDSAT 5 images in 8-bit
unsigned integer, 16-bit signed integer, 16-bit unsigned integer
and 32-bit floating-point real image channels. The 8-bit data
format is designated as 8U and this format is most often used
for air and satellite imagery (PCI Enterprises, 1998, pp. 22,27).
The spectral range of each LANDSAT 5 TM band that is set to
the 8-bit format is converted to 256 possible gray levels
starting at 0 and ending at 255. The levels are numbered by
row and column. Each of these pixels represents a particular
light reflectance characteristic of the feature (or features) it
represents on the earth’s surface. Depending on the size of an
object on the earth’s surface, it is possible that a LANDSAT 5
TM pixel may be representing more than one ground or water
feature and this can affect the reliability of the reflectance
value of the pixel.
LANDSAT 5 TM (Thematic Mapper) bands 1, 3, 4, and 5 were
used to create the three elements of the coastal zone since these
bands have spectral ranges that help in identifying the water
and land interface. The wavelengths and characteristics of
these LANDSAT 5 TM bands as summarized from Lillesand
& Kiefer (1994, p. 468) are as follows:
•
•
•
•
Band 1 has a wavelength of 0.45 to 0.52 µm and it has a
nominal spectral location of blue. Band 1 can penetrate
water, which makes it appropriate for mapping coastlines.
Band 1 is also useful for the identification of soils and
plants including forests, as well as cultural features.
Band 3 has a wavelength of 0.63 to 0.69 µm and it has a
nominal spectral location of red. Band 3 can detect
chlorophyll, which aids in plant identification.
Band 4 has a wavelength of 0.76 to 0.90 µm and its
nominal spectral location is the near infrared. Band 4 is
useful for interpreting different types of vegetation,
detecting moisture in soils and for the delineation of water
and land.
Band 5 has a wavelength of 1.55 to 1.75 µm and its
nominal spectral location is the mid-infrared. Band 5 can
be used to detect the moisture content of various plants
and soils.
By arranging these LANDSAT 5 TM bands in various
combinations with Image Handler, different features of the
coastal zone were easy to recognize as specific colors such as
dark blue for water, reddish tones for tidal flats and various
greens for the islands and the mainland. These different color
combinations made it much easier to perform a visual
identification and classification of some of the more important
coastal zone features. Many features on the earth’s surface
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
including vegetation, soils, rocks, and water have a reflectance
value that is visually more noticeable in some LANDSAT 5
TM bands than in the other bands (the interpretation of images
using remote sensing often requires the human eye to aid in
identifying and classifying a feature on the earth’s surface).
Using a program such as Image Handler, the various pixels that
are contained within the converted LANDSAT 5 TM image
were reclassified into three zones of water, land and intertidal.
This method was chosen since these three zones were
relatively easy to classify based on their spectral values. Then
the water and intertidal zones were joined to create a zone that
marks the MHW (mean high water) or high tide zone. All of
the ground features that made up the mainland such as trees,
rocks, roads, airport runways, docks, bridges, wharfs, buildings
and grass have different spectral characteristics, but since it is
only necessary to identify the land as a whole, these elements
can be combined into a single and larger group of pixels that
are designated with one value. This new single group of pixels
has a spectral range that represents features that are associated
with the landmass. The same technique was used on pixels that
represented the spectral characteristics of the water areas as
well as those pixels that represented the features in the
intertidal zone. Once the original LANDSAT 5 TM image is
converted to a two-zone image it is then saved as a new layer
within the PCI Image Handler, and this new layer can then be
prepared for use in a geographic information system where the
coastal zone features can then be mapped.
8.5 Step Four: Loading the GeoTIFF File into ARCVIEW
for Spatial Analysis
The PCI file was exported as a GeoTIFF image into ArcView
for the purpose of being a georeferenced image for the onscreen digitizing of the shoreline of the mainland and of the
islands of Cobscook Bay. A new ArcView project was then
created that used the same cartographic specification as the
geographic projects had in Case Study A. The map and
measuring units were set to meters. The map datum for this
project was set as NAD83, which is essentially the same map
datum as WGS84 for the original LANDSAT 5 TM image.
The GeoTIFF file was used as a base image so that the
digitizing of shorelines and the creation of polygons for water
areas could be accomplished within ArcView. Then, as was
done with the digital spatial data in Case Study A, the length
and areas of features were calculated based on the three extents
of Cobscook Bay (Table 3).
TABLE 3. LANDSAT 5-BASED ANALYSIS OF COBSCOOK BAY (KOSTIUK, 2001A).
Dimension or count.
Comstock
Point
Shackford Head
Length of Mainland
Shoreline. Meters
to
Comstock Point to Estes
Head
Estes Head to Lubec
Town Pier
271,208
276,142
288,489
Number of Islands.
41
41
46
Length
of
Island’s
Shorelines. Meters
39,762
39,762
45,202
Area of Islands. Square
Meters
2,624,631
2,624,631
2,937,095
Area of Cobscook Bay.
Square Meters
81,147,847
83,289,246
91,275,792
8.4 Checking the Horizontal Accuracy of the Spatial
Analysis of Case Study B
As was demonstrated in Case Study A the horizontal accuracy
of spatial data can vary considerably by scale and data source.
The values in table 3 show the results of using remote sensing
data with a GIS to obtain measurements of length and area.
What is lacking in the spatial analysis of case study B is an
assessment of the accuracy of the location of the shoreline. In
order to determine the horizontal accuracy of the geographic
representation of the shoreline of Cobscook Bay that was based
on LANDSAT 5 TM data a new GIS project was set up. The
1:24,000 scale vector data of Cobscook Bay was used to
reference the horizontal position of the shoreline that was
created from the LANDSAT 5 TM data. Referring to the
metadata that was written by the USGS, the horizontal
accuracy of the 1:24,000 data is 40 feet or 12.19 meters on the
ground (USGS, Part 1, 1997, p.1. D-2). Therefore, the level of
inaccuracy that is present in the 1:24,000 data limits the
measurement of the horizontal accuracy of the shoreline that
was created from the LANDSAT 5 TM data to 12.19 meters.
Fifty control points were established at random locations
across Cobscook Bay so that the position of the 1:24,000-based
and LANDSAT 5 TM-based shorelines could be measured and
recorded. The difference in the location of the two shorelines
at the fifty control points were measured by vector lines and
the length of the line segments were calculated in meters using
the calcapl script. An example of the horizontal offset between
the 1:24,000-based and the LANDSAT 5 TM-based shorelines
are shown in figure 1. The 1:24,000 shoreline is dotted while
the LANDSAT shoreline is a solid line.
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
Figure 1
The statistics function of ArcView was used to calculate
various elements of the horizontal accuracy of the LANDSAT
5 TM -based shoreline and they are shown below:
•
Count: 50 (the number of control points used for
measuring the offset of the two shorelines).
•
Mean: 58.327 (he average distance in meters of the offset
of the two shorelines).
•
Maximum: 314.301 (the largest offset between the two
shorelines).
•
Minimum: 5.811 (the smallest offset between the two
shorelines).
•
Range: 308.490 (the difference between the largest and
smallest offset).
•
Standard Deviation: 54.697 (the standard deviation of the
offsets).
The horizontal accuracy of the shorelines that were created in
case study B ranged from 5.8 meters to 314.3 meters with the
average horizontal accuracy being 58.3 meters. Referring back
to the rule of detection and resolution size that was described
in earlier (see table 1), the minimum detection size of case
study B will be 60 meters since 30 meters is the resolution size.
Since the horizontal error is 58.3 meters this is an acceptable
(and expected) error because it conforms to the detection and
resolution formula where the detectable unit is twice that of the
resolution unit (Tobler, 1988). The horizontal accuracy of 58.3
meters also is a validation of the methods used to identify the
shoreline in case study B since the error fell within the
expected limits according to the detection and resolution rule.
However, since the detection size is 60 meters no apparent
change of the coastline can be attributed to the analysis of
Cobscook Bay by the LANDSAT 5 data. This is because the
measurements of the horizontal location of the shorelines
between Case Study A and Case Study B are within the
expected error rate of the detection grid. Therefore, only
remote sensing results that are greater than the size of the
detection grid should be used to measure sea level and
shoreline change.
Knowing the level of accuracy of spatial data is obviously
useful for particular coastal zone management applications
where the actual geographic position of the coastline is
important. For example, horizontal accuracy of spatial data are
important for the analysis and mapping of shoreline erosion,
detecting shoreline change, tracking and clean up of oil spills,
habitat assessment and mitigating damage due to storm surges.
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
9.
DISCUSSION OF THE SPATIAL ANALYSIS OF
CASE STUDY B
The results that were obtained by using the LANDSAT data
were quite different from the results that were obtained by
using the 1:24,000 scale digital spatial data in Case Study A.
For example the length of the shoreline of Cobscook Bay that
used the Lubec town pier limit showed a value of 288,489
meters for the LANDSAT 5-based data while the 1:24,000
spatial data has a result of 380,900 meters. This was a
difference of 92,411 meters and this clearly shows that
LANDSAT 5 TM images cannot be relied upon to map a
shoreline feature at the same resolution as is displayed on
1:24,000 spatial data. Since the minimum acceptable scale for
using digital spatial data for use in a geographic information
system to measure coastline features was determined by
Wainwright (1991) to be 1:40,000 the results that were
obtained from the LANDSAT 5 TM-based spatial analysis
clearly shows that it does not meet these requirements. The
LANDSAT 5 TM -based spatial analysis also did not produce
results that are as good as those that were produced from the
1:100,000 scale digital spatial data.
There are several reasons why the LANDSAT 5 TM images do
not meet the minimum requirements, and the two main reasons
are scale and resolution. The scale of the LANDSAT 5 TM
image is approximately 1:1,000,000 and the resolution of the
pixel size is 30 meters x 30 meters which limits how much
detail can be clearly seen and/or interpreted correctly. When
the LANDSAT 5 TM images were used for spatial analysis
with both the PCI and ArcView systems, the ability to see and
recognize fine ground detail such as residential streets and
small buildings was limited. Personal knowledge and
experience of the study area made it slightly easier to recognize
certain features, and this helped in determining where such
features as the normal high and low tide lines were located.
LANDSAT 5 TM data have limited resolution and it is only
appropriate to provide a regional view. If remote sensing data
are to be used for effective coastal zone management purposes
such as measuring coastline features then the image pixels
must provide a ground resolution much larger than 30 meters
(such as 4 meters). There are such data available and one
source is provided by the new IKONOS remote sensing
satellite. For example, the IKONOS data has a precision of 1meter panchromatic or 4-meter multi-spectral imagery. Using
the rule of the sampling interval would mean that IKONOS
Remote sensing data should have a detection accuracy of 2
meters for the panchromatic (1-meter) and 8 meters for the
multi-spectral imagery (4 meters). Another reason for the
difference in the measurements of the features in Cobscook
Bay is due to the boundary that was arbitrarily set for the
interior limits of the bay. The location of the coastline on the
1:24,000 scale data was taken from USGS maps which used
the Mean High Water (MHW) line and “the extent of tidal
features was determined by a group of marine specialists
familiar with Maine's coast (Maine Office of GIS, 2000).
Unfortunately, no such methods of ground verification were
used for the case study. Consequently, during the process of
on-screen digitizing for case study B there were varying
degrees of difficulty to determine where the coastline stopped
and the rivers began. The areas where the inner limits of
Cobscook Bay were assumed to be located may have been, in
many cases, premature. This type of error would mean that a
shorter shoreline and a smaller water area would be measured
during the spatial analysis stage. This problem is one of the
yet to be defined questions of the coastal zone such as: Where
does the coastline, or coastal zone begin or end?
10. CONCLUSION
Detecting sea level change requires accurate data and
information, and Remote Sensing data can have an important
role to play in this discipline. In order for Remote Sensing data
to be able to play an important and vital role for coastal zone
management purposes there must be a set of minimum
acceptable standards for the use of spatial data (Standards like
those that formed the basis of case studies A and B are a good
example). The type and scale of spatial data that are used in
geographic information systems must match a set of
cartographic requirements and conditions in order for them to
be considered as being reliable and useful. Therefore it is
important to know what level of precision and accuracy is
required for every coastal zone management application that
needs the use of spatial data. Acquiring the image at a point in
time that matches the selected Vertical datum will also ensure a
more reliable result as well as reducing the time required to
perform the spatial analysis.
Remote sensing data such as LANDSAT 5 TM satellite images
can be useful for reconnaissance purposes to check if there has
been significant change in a coastline or to detect various types
of environmental conditions such as storm surges, sediment
transport along the coast, or to track the spread of oil from
ships or land based sources. If the detection unit of the spatial
data (twice the resolution unit) matches the scale that is being
used for an application then the spatial data can be used for
direct updating of maps or digital spatial data. If the spatial
data do not meet minimum acceptable requirements then more
accurate sources and images should be used in geographic
information systems. Management and analysis of the coastal
zone require both an accurate and defensible spatial analysis of
various coastal features. Remote sensing data have a
potentially valuable role to play in this field especially as the
scale and the resolution of the images improve over time.
REFERENCES
ArcView GIS 3.2. Computer Software. ESRI (Environmental
Systems Research Institute, Inc.) (1999).
Brooks, D. A., Baca, M. W., & Lo, Y. T. (1999). Tidal
Circulation and Residence Time in a Macrotidal Estuary:
Cobscook Bay, Maine. Estuarine, Coastal and Shelf Science
49, 647-665.
Environmental Systems Research Institute, Inc. (1993). The
Digital Chart of the World for use with ARC/INFO. Data
Dictionary. Redlands, CA: Author
Gaudet, R. J. (1990). Mapping the Coastal Zone - Can One
Base Satisfy Many Issues? GIS for the 1990s. Proceedings of
the Second National Conference in Geographic Information
Systems. pp 1357-1361. Ottawa.
GeoInformation International. (1997). Getting to Know
ArcView GIS. The Geographic Information System for
Everyone. Cambridge: Author.
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
Geomatics Canada. (1997). Standards and Specifications of the
National Topographic Data Base. (ed. 3.1) Sherbrooke
(Québec): Minister of Supply and Services Canada.
Jensen, John, R. (1996). Introductory Digital Image
Processing. Upper Saddle River, NJ: Prentice Hall, Inc.
Kostiuk, M. (2001a). THESIS: Determining the Spatial Data
Requirements for a GIS to Support Coastal Zone Management.
Cobscook Bay, Maine. A Case Study. Ottawa: Carleton
University.
Kostiuk, M. (2001b). The Use of Geographic Information for
Coastal Zone Management. Cobscook Bay, Maine - A Case
Study. In T. Chopin & P.G. Wells (Eds.), Opportunities and
Challenges for Protecting, Restoring and Enhancing Habitats
in the Bay of Fundy. Proceedings of the 4th Bay of Fundy
Science Workshop, Saint John, New Brunswick, September 1921, 2000. Environment Canada, Atlantic Region Occasional
Report Number 17, Environment Canada, Dartmouth, Nova
Scotia
Lillesand, T. M., & Kiefer, R. W. (1994). Remote Sensing and
Image Interpretation (3rd ed.). New York: John Wiley & Sons,
Inc.
Maine. Office of Geographic Information Systems (2000).
Data Catalog and MetaData, Coast: Mean high water coastline
from USGS 1:24,000 scale quads, Coast Coverage Information
[Online].
Available: http://apollo.ogis.state.me.us/catalog/catalog.htm
Monahan, D. (2000). Zones in the Sea. From Shoreline to Shelf
Edge. A Workshop on Article 76 of The United Nations
Convention on Law of the Sea (UNCLOS) for Participants in
CZ2000 Version Zero (0). Canadian Hydrographic Service &
Department of Geodesy and Geomatics Engineering,
University of New Brunswick. [On-Line] Also Available:
ftp.chs-shc.dfo-mpo.gc.ca/Outgoing/MonahanD/Coastal.
Rohmann, Steve.(August, 2000) NOAA's Medium Resolution
Digital Vector Shoreline.
[On-Line]. Available:
http://seaserver.nos.noaa.gov/projects/shoreline/shoreline.html
Soluri, E. A., & Woodson, V.A. (1990). World Vector
Shoreline. International Hydrographic Review, [On-Line],
LXVII (1). Available: http://crusty.er.usgs.gov/coast/wvs.html
Tobler, W. (1988). Resolution, Resampling, and All That. In
Mousey , H. &R. F. Tomlinson (Eds.), Building Databases for
Global Science. The Proceedings of the First Meeting of the
International Geographical Union Global Database Planning
Project. (pp.129-137). London: Taylor and Francis.
USGS (United States Geological Survey). (1997). Part 1
Standards for 1:24,000-Scale Digital Line Graphs-3 Core Part
1: Data Description and Template Development National
Mapping Program Technical Instructions. United States of
America: U.S. Department of the Interior. U.S. Geological
Survey. National Mapping Division.
USGS (United States Geological Survey). (1997). Part 3
1:100,000-Scale Digital Line Graphs Standards for the
Preparation of Digital Geospatial Metadata. National
Mapping Program Technical Instructions. United States of
America: U.S. Department of the Interior. U.S. Geological
Survey. National Mapping Division.
USGS and NOAA. (2002). NOAA and USGS Tampa Bay
Demonstration Project Area VDatum Transformation Tool.
[On-Line]
http://chartmaker.ncd.noaa.gov/bathytopo/vdatum.htm (August
1, 2002).
Wainwright, Peter. LGL Limited.
Canada Department of
Fisheries and Oceans; Fish Habitat and Information Program
(Canada). (1991). Fisheries Habitat GIS Strategy. Vancouver:
LGL Ecological Research Associates.
USING REMOTE SENSING DATA TO DETECT SEA LEVEL CHANGE
Pecora 15/Land Satellite Information IV/ISPRS Commission I/FIEOS 2002 Conference Proceedings
Download